prey selection and kill rates of … investigated prey selection and kill rates of cougars in...
TRANSCRIPT
PREY SELECTION AND KILL RATES OF COUGARS
IN NORTHEASTERN WASHINGTON
By
HILARY STUART CRUICKSHANK
A thesis submitted in partial fulfillment of the requirements for the degree of
MASTER OF SCIENCE IN NATURAL RESOURCE SCIENCES
WASHINGTON STATE UNIVERSITY Department of Natural Resource Sciences
DECEMBER 2004
ii
To the Faculty of Washington State University: The members of the Committee appointed to examine the thesis of HILARY STUART CRUICKSHANK find it satisfactory and recommend that it be accepted.
______________________________________ Chair
______________________________________
______________________________________
iii
ACKNOWLEDGEMENTS
This thesis is the product of many generous collaborators and supporters. Funding and
in-kind support were provided by Bonneville Power Administration Fish and Wildlife
Program, Washington Department of Fish and Wildlife, U.S. Forest Service, Washington
State University, and Bombardier, Inc.. I would like to thank my advisor, Dr. Robert
Wielgus, for his confidence in me, accessibility, and encouragement. Dr. Lisa Shipley
and Dr. Charles Robbins provided support, enthusiasm, and scientific expertise. Heartfelt
thanks to Hugh Robinson, my “surrogate advisor” and “normal” friend, who taught me
just about everything I know about cougars. I would also like to thank the generous
efforts of all who contributed in the field, including Catherine Lambert, Dave Parker,
Tom MacArthur (along with Newly, Sooner, and Striker), Dan Ravenel, Nikki Heim,
Crystal Reed, Kate Odneal, and Emma. Thanks to my family, who continuously provide
me encouragement and support in all that I do. Finally, and most importantly, to Skye
Cooley, for his patience, loving support, technical expertise, and endless creative ideas. I
couldn’t have done it without you.
iv
PREY SELECTION AND KILL RATES OF COUGARS IN NORTHEASTERN WASHINGTON
ABSTRACT
by Hilary Stuart Cruickshank, M.S. Washington State University
December 2004 Chair: Robert B. Wielgus
We investigated prey selection and kill rates of cougars in northeastern
Washington from 2002-2004, in a sympatric white-tailed deer and mule deer system. We
tested two competing hypotheses of prey selection, “prey switching” and “apparent
competition”. We developed a sightability model which corrected ground counts of
white-tailed deer and mule deer using life-sized deer decoys to calculate relative prey
availability. A logistic regression sightability model tested for effects of group size,
distance, and habitat on deer sightability, then predicted relative numbers (availability) of
both deer species on transects. To estimate use of prey by cougars, we examined 60
cougar kills. White-tailed deer comprised 60% of the kills (mule deer comprised 40%), a
proportion larger than the study area’s prey population (70% white-tailed deer vs. 30%
mule deer). Cougars selected for mule deer across the entire study area. We also
detected strong seasonal changes in prey selection, with cougars strongly selecting for
mule deer in summer, but not during winter. Mean annual kill rate was 6.68 days per
deer killed. Kill rates did not differ between seasons or deer species. Habitat
characteristics of kill sites did not differ significantly between white-tailed deer and mule
deer kills. These findings are consistent with the apparent competition hypothesis and
v
suggest that the current decline in mule deer numbers in northeastern Washington is
caused by an abundant invading primary prey species (white-tailed deer) and a related
increase in predation on the secondary prey species (mule deer) during summer months.
vi
TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS………………………………………………………………iii
ABSTRACT……………………………………………………………………………...iv
LIST OF TABLES……………………………………………………………………...viii
CHAPTER 1: ESTIMATING PREY AVAILABILTIY IN HOME RANGES WITH A
SIGHTABILITY MODEL………………………………………………………………….
ABSTRACT…………………………………………………………………….....1
INTRODUCTION………………………………………………………………...2
STUDY AREA…………………………………………………………………....5
METHODS………………………………………………………………………..7
MODEL DEVELOPMENT……………………………………………….7
GROUND SURVEYS OF LIVE DEER………………………………….9
HELICOPTER SURVEY………………………………………………..10
MODEL COMPARISONS………………………………………………11
RESULTS………………………………………………………………………..12
SIGHTABILITY FACTORS…………………………………………….12
GROUND COUNT……………………………………………………....12
HELICOPTER SURVEY………………………………………………..13
MODEL COMPARISONS………………………………………………14
DISCUSSION……………………………………………………………………15
vii
LITERATURE CITED…………………………………………………………..18
CHAPTER 2: PREY SELECTION AND KILL RATES OF COUGARS IN
NORTHEASTERN WASHINGTON……………………………………………………27
ABSTRACT……………………………………………………………………...27
INTRODUCTION…………………………………………………………….....28
STUDY AREA…………………………………………………………………..30
METHODS………………………………………………………………………32
COUGAR KILLS………………………………………………………..32
PREY AVAILABILITY…………………………………………………33
PREY SELECTION AND KILL RATES…………………………….....35
HABITAT CHARACTERISTICS AT KILL SITES……………………37
RESULTS………………………………………………………………………..39
COUGAR KILLS………………………………………………………..39
PREY AVAILABILITY…………………………………………………39
PREY SELECTION AND KILL RATES……………………………….40
HABITAT CHARACTERISTICS AT KILL SITES……………………42
DISCUSSION……………………………………………………………………43
LITERATURE CITED…………………………………………………………..46
viii
LIST OF TABLES
Table 1.1 Process of model selection for a logistic regression analysis of deer sightability
in northeastern Washington, 2002-2004……………………………………..23
Table 1.2. Coefficients, standard errors, probabilitites of significance (t-ratio), and odds
ratios for group size, distance and habitat parameters used in a logistic
regression model for deer sightability in northeastern Washington during fall
2003………………..…………………………………………………………24
Table 1.3. Raw (obs) and corrected (cor) numbers of deer observed in each of four
habitat types from ground surveys in 2003 for northeastern WA……….…..25
Table 1.4. Estimated total (N) and relative (%) numbers of white-tailed deer and mule
deer from corrected ground and aerial surveys in 2002-2004 in northeastern
WA…………………………………………………………………………...26
Table 2.1. Chi-square test results and selection ratios for 2nd order cougar prey selection
in northeastern Washington, during 2002-2004. Prey availability within study
areas, observed deer kills, and expected deer kills is given for white-tailed
deer (WT) and mule deer (MD). Omnibus statistics show results of pooling
across areas and cougars………………………………..……………………50
Table 2.2. Chi-square test results and selection ratios for seasonal 2nd order cougar prey
selection in northeastern Washington, during 2002-2004. Prey availability
within cougar home ranges, observed deer kills, and expected deer kills is
given for white-tailed deer and mule deer…………………………………...51
ix
Table 2.3. Continuous variables from habitat characteristics at kill sites of white-tailed
deer and mule deer investigated during 2002-2004 in northeastern
Washington…………………………………………………………………..52
Table 2.4. Categorical variables from habitat characteristics at kill sites of white-tailed
deer and mule deer investigated during 2002-2004 in northeastern
Washington…………………………………………………………………53
1
CHAPTER 1
ESTIMATING PREY AVAILABILITY IN HOME RANGES WITH A
SIGHTABILITY MODEL
ABSTRACT
This study introduces an inexpensive, and convenient method to estimate ungulate
sightability and relative numbers from ground counts using life-sized animal decoys.
Logistic regression (LR) was used to test for effects of group size, distance, and habitat
on deer sightability. The best-fit LR model was then applied to observations of free-
ranging mule deer and white-tailed deer on transects to calculate their relative numbers.
The model estimated relative populations in our northeastern Washington study area at
72% white-tailed deer and 28% mule deer. We compared this model to results obtained
through an aerial count, and also tested for seasonal and geographic differences in
relative prey availability. The corrected ground count and the aerial count yielded
statistically equal proportions of deer. We detected significant seasonal and geographic
variation in the ratio of deer species. When averaged across seasons, sub-areas, and
surveys, relative prey availability was 70% white-tailed deer and 30% mule deer.
2
INTRODUCTION
Researchers must accurately determine animal numbers to meet wildlife
management objectives. However, animal counts are often biased because seldom are all
animals seen during surveys. When conducting surveys, observers count some unknown
fraction of the total animals present. This fraction must be translated to estimates of
population size. Sighting probability, or sightability, is the probability that an animal will
be seen by an observer during a survey (Krebs 1999). Several methods are available to
estimate sightability for animal counts including double sampling, marked sub-sample,
line transects, quadrat counts, removal methods, capture-recapture, and plotless methods
(Anderson 2003). Most of these methods are expensive (e.g., some require radio-collared
animals), time consuming, and are often impractical. This study introduces an
inexpensive and convenient method to estimate sightability for ungulate ground counts
using life-sized animal decoys as “marked” animals.
To produce estimates that correlate with the actual population size, counts must
relate to sightability. Many variables affect an animal’s sightability such as observer
effects (experience, interest, eyesight, fatigue), environmental variables (precipitation,
habitat, time of day, vegetation type, temperature), and aspects of the species (color,
behavior, group size) (Anderson 2001). These variables change depending on when,
where, and what a researcher surveys, thus chances of observing an animal will also
change with time, place, and target animal. Unless one can estimate the detection
probability to relate count data (index data) to the size of the true population, one cannot
assume it is representative of the population. The true population size is related to the
3
index or count value (C) as N = C/p, where p is the detection probability of the animal
being observed, or sightability (Anderson 2003).
Sightability models allow estimates of detection probabilities, thereby correcting
for animals not seen during surveys. Sightability models establish detection probabilities
by incorporating variation in detection among habitats, seasons, years, species, and
distances (Williams et al. 2002). Once the detection probability (^p ) is known, the
parameter, ^^
/ pcN = , can be calculated, and a confident population estimate made.
Logistic regression (LR) estimators are often used to develop sightability models
because they can incorporate categorical and continuous, non-parametric, non-additive,
and non-linear independent variables to predict a binomial dependent variable (sighted,
not sighted) (Kleinbaum et al. 1982). LR models help eliminate problems associated
with heterogeneous sightability among animals by correcting for each group of animals
observed. Unlike mark-resight methods, marked animals are only needed during model
development (Bartmann et al. 1987). Because of their flexibility and effectiveness,
researchers have used LR models for aerial surveys of elk (Cervus elaphus) (Samuel et
al. 1987, Otten et al. 1993), mule deer (Odocoileus hemionus) (Ackerman 1988), and
moose (Alces alces) (Anderson 1996). While this method has proven successful for
aerial surveys, researchers have not yet applied it to ground counts.
As part of an ongoing project studying cougar prey selection in northeastern
Washington, we needed a simple, effective, and inexpensive way to assess differences in
prey availability. While tracking cougars in the field, we conducted ground counts of
white-tailed deer and mule deer during June 2002 – March 2004. Our goal was to
develop and validate a technique that used ground counts to infer prey availability over
4
space and time. To correct the ground counts for sighting biases, we developed a
sightability model using LR and 6 deer decoys set in 48 combinations. We tested for
differences in sightability between the two deer species in our study area. Although mule
deer and white-tailed deer are similar in size and color, they use different habitats and
display different behaviors, which could result in different sightabilities.
Specifically, our objectives were to 1) identify environmental variables that affect
sightability of white-tailed deer and mule deer during year-round ground counts, 2)
develop a sightability model to correct for biases in ground counts, 3) validate the model
by comparing results with a winter helicopter survey, and 4) determine annual and
seasonal relative abundances of white-tailed deer and mule deer for radio-collared
cougars.
5
STUDY AREA
The study area covers approximately 3,800 km2 in northeastern Washington.
Boundaries extend from the Okanogan/Ferry County line west to the Columbia River,
and from the Canadian/US border south to the Colville Indian Reservation. The study
area lies entirely within the Okanogan Highlands physiographic province, composed of
glacially subdued mountainous terrain, with elevations ranging from 400 m in the
Columbia River valley to 2130 m at the top of the Kettle Crest. Forest overstory species
include Douglas-fir (Pseudotsuga menziesii), western hemlock (Tsuga heterophylla),
western red cedar (Thuja plicata), and subalpine fir (Abies lasiocarpa). Inland maritime
conditions characterize the climate, with mean temperatures ranging from –6 °C in
January to 21 °C in July, and annual precipitation of 46 cm. Snowfall averages 136 cm
during a 6-month period from mid-November to mid-April.
Field efforts were concentrated in two separate areas separated by the Kettle
River. “The Wedge” study area lies to the east of the Kettle River, and the “Republic”
area lies to the west of the River and the Kettle Crest Mountain Range. We conducted
analyses on each replicate study area separately.
Robinson et al. (2002) found that white-tailed deer (Odocoileus virginianus) were
the most abundant ungulate in a nearby study area, followed by mule deer (Odocoileus
hemionus). Since climate and physiography create seasonally migratory deer
populations, both white-tailed deer and mule deer congregate on winter ranges between
December and April. Deer winter ranges are generally on south to west-facing gentle
slopes in timber stands with higher canopy closure, providing wind and snow breaks
6
(Pauley et al. 1993, Armeleder et al. 1994). Across the study area, higher elevation
winter ranges (i.e., 900-1200 m) are almost exclusively occupied by mule deer, whereas
lower elevation ranges (i.e., ≤ 900 m) are predominantly used by white-tailed deer.
During summer white-tailed deer move up in elevation and intermix with mule deer.
7
METHODS
Model Development
To determine detection probabilities for deer we conducted sightability trials,
which standardized several controllable factors (e.g., number and experience of
observers, driving speed), and measured biases of sightability caused by environmental
factors (e.g., group size, distance, vegetation cover). Beginning in August 2003, 6
McKenzie HD30 deer decoys (McKenzie Targts, Granite Quarry, N.C.) were placed
daily along secondary roads in the study area. Decoys were placed in groups of 1 to 6
individuals, in 1 of 4 habitat types (open ponderosa pine, dense mixed forest, clear-cut,
and agricultural), from 0 to 130 m from transects. A second researcher drove the road
later that day and recorded the decoys observed.
We used logistic regression (Proc Logistic, SAS Institute, Cary N.C.) to estimate
the relative importance and parameter values of the independent variables (group size,
distance to deer, and habitat type) on the dependent variable (individual deer sightings
and non-sightings). We examined the influence of each variable on deer sightability by
determining the effectiveness of the full model against models with a reduced number of
parameters; in effect testing the null hypothesis that group size, perpendicular distance,
and habitat type did not influence deer sightability (Ott and Longnecker 2001). We used
the VIF (Variation Inflation Factor) diagnostic tool in SAS (SAS Institute, Cary, N.C.) to
test for collinearity between independent variables.
To correct for individual sightability, we considered each decoy as an individual
observation because decoys could be in groups ≥ 1. Steinhorst and Samuel (1989)
8
suggested that if sightability is constant then success in observing an animal can be
viewed as a simple binomial experiment (sightings and non-sightings). Given this, we
coded the dependent variable, whether an animal was sighted or not, as 0 for sighted and
1 for not sighted. When a group of two or more decoys was placed in the forest, each
decoy was assessed for individual sightability. For example, if only one of the decoys
was observed, we categorized one as a success (coded as a 0), and one as a failure (coded
as a 1). Although individuals in a group may seem to violate the assumption of
independence between sampled units, our goal was to develop a correction factor for
individual sightability with group size as an independent variable (Krebs 1999); so a
violation of independence was not relevant in this case.
The LR model for sightability was:
µ
µ
eep+
=∧
1,
where ∧
p is the detection probability, e is the natural log 2.718, and µ= β0 + β1x1 + β 2x2 +
...+ βkxk is the set of variable parameters (β) multiplied by the independent variables (x1,
x2,...,xk) (Unsworth et al. 1999). We corrected for each sighting of decoys based on the
equation:
^^/ pCN = ,
where C is the original count or index data, and ^N is the corrected population estimate.
9
Ground Surveys of Live Deer
To determine prey available to each radio-collared cougar within the study area,
we did not establish static sampling transects, but instead adapted a dynamic sampling
technique to follow the cougar’s movements throughout their home ranges. From 2002-
2004, researchers monitored radio-collared cougars’ movements through daily, year-
round ground telemetry. While conducting field work researchers documented all deer
encountered in the cougars’ home ranges on foot or in a vehicle. Because of the
repetitious commute to our study area on a paved highway along the Columbia River, we
omitted all deer observations on this highway for calculation of prey availability. On
occasions when transects (roads, trails) were covered more than once per day, we
recorded deer sightings only once per day. Because we were interested in relative
numbers of available prey, rather than absolute availability, we avoided any biases of
pseudo-replication that might occur by recording the same deer several days in a row.
At each observation, the date, time, Universal Transverse Mercator (UTM)
coordinate, species, number of deer, habitat, and the straight-line distance from the
transect were recorded. We classified deer as male or female, and adult or fawn. Habitat
types were classified as open ponderosa pine, dense mixed forest, clear-cut, or
agricultural.
Of these variables, we identified three that had the highest potential to affect the
probability of detection for each deer species, and therefore the accuracy of our index:
group size, habitat type, and perpendicular distance from the transect (road or trail)
(Buckland et al. 2001). We examined whether sightability was the same for white-tailed
deer and mule deer by testing for differences in independent variables between species.
10
We used the t-test (Proc Ttest, SAS Institute, Cary N.C.) to test for differences in group
size and perpendicular distance among deer species. We conducted a chi-square test of
homogeneity (Proc Freq/Chisq expected, SAS Institute, Cary N.C.) to test for a
species/habitat association.
Helicopter Survey
As a second method of determining relative deer abundance, we conducted a
helicopter survey on The Wedge portion of our study area. Because of funding
restrictions we surveyed half of the study area for one year only. We identified 20
subunits with clear boundaries distinguishable from the air prior to conducting the
survey. Size of subunits ranged from approximately 4 km2 to 17 km2 and each required <
1 hour to survey. We used ground count data, along with knowledge of the district
biologists to define and stratify subunits as low or high quality white-tailed deer and mule
deer habitat. We then selected 5 subunits for each of the 4 strata to survey according to
methods described by Unsworth et al. (1999).
In February 2004, we surveyed all 20 subunits to determine relative white-tailed
deer and mule deer availability. We flew each subunit via transects 200-500 m apart and
at a consistent speed range of 65-80 km/h. Two researchers in the back seat observed
from both sides of the helicopter while a third researcher recorded observations. When
deer were spotted, the helicopter paused and circled the area to confirm the observation.
We recorded group size, activity of the animal, canopy cover, habitat, and percent snow
cover. The same primary observer and pilot were used throughout the survey, while
secondary staff varied from day to day. Point estimates, variances, and confidence
11
intervals were calculated based on formulas in the Mule Deer Spring model in program
AERIAL SURVEY version 1.0 (Unsworth et al. 1999).
Model Comparisons
To validate the effectiveness of our ground survey, we compared the proportion
of white-tailed deer and mule deer from the corrected winter ground survey on The
Wedge to that from the aerial survey. We also compared seasonal vs. annual proportions,
and proportions of deer species in the entire study area vs. the two sub-areas (The Wedge,
Republic). We used the chi square test of homogeneity for all comparisons.
12
RESULTS
Sightability Factors
From sightability trials using deer decoys, we were most successful at observing
deer in open habitats, in larger group sizes, and at shorter distances. Success rates for
individual deer sightings were 0.93 for agricultural areas, 0.80 in open ponderosa pine
forest, 0.70 in clear cuts, and 0.23 in dense mixed forest. We observed 0.50 of deer in
group sizes of 1 animal, 0.53 of deer in group sizes of 2 animals, and 0.76 of deer in
group sizes of 3 + animals. We also observed 0.96 of deer at 0-50 m, 0.45 at 50-100 m,
and 0.48 at 100 +m.
Sightability of decoys depended on habitat, group size, and distance. The VIF
diagnostic showed little collinearity between independent variables (VIF < 1.01). The
sightability models are given in Table 1.1, and parameter values for the best model are
given in Table 1.2. No additional 2-way interaction models yielded significant chi-square
improvements. The -2 Log Likelihood for the best model was 72.391 (d.f. = 5, p = 0.00)
(Manly et al. 1993), and McFadden’s Rho-Squared was very good at 0.476 (Hosmer and
Lemeshow 1989) (Table 1.1).
Ground Count
We followed 15 radio-collared cougars from June 2002 through December 2003.
Raw prey availability counts indicated that white- tailed deer were more abundant than
mule deer, with more groups (317 vs. 150) and total individuals (843 vs. 355) observed
13
during daily field work. The uncorrected relative deer population across the study area
was 70% white-tailed deer and 30% mule deer.
We observed more white-tailed deer in open agricultural areas and at longer
distances than mule deer; and observed mule deer more often in forested habitats closer
to the observer (Table 1.3). We found a significant species by habitat association for
individuals (χ2 = 36.58, d.f. = 3, p < 0.001) and groups of deer (χ2 = 24.63, d.f. = 3, p <
0.001). We also found a significant difference in distance between species (t = 3.49, d.f.
= 370, p < 0.001), with mean distance and standard error for white-tailed deer at 23.6 ±
33 and 13.9 ± 25.4 for mule deer. We found no significant difference in group size (t =
0.89, d.f. = 392, p = 0.37). Mean group size and standard error for white-tailed deer was
2.66 ± 4.05 and 2.37 ± 2.91 for mule deer.
When applied to ground observations, the sightability model increased the
estimated number of white-tailed deer 191% from 843 to 1612, and increased the mule
deer estimate 180% from 355 to 639. Corrected relative availability across the entire
study area was 72% white-tailed deer and 28% mule deer.
Helicopter Survey
The February 2004 aerial survey on the wedge indicated that white-tailed deer
were more abundant than mule deer. The Spring Mule Deer model (Unsworth et al.
1999) estimated population size and 90% confidence intervals at 1,384 ± 221 (80%) for
white-tailed deer and 354 ± 83 (20%) for mule deer. As expected, sightability variance
accounted for nearly all of the total variance for both white-tailed deer (84.6%) and mule
deer (89.5%).
14
Model Comparisons
The ground survey proved to be an effective method to measure relative
abundances of white-tailed and mule deer populations. We found no difference (χ2 =
0.0002, d.f. = 1, p = .99) between the ground survey and helicopter survey on The Wedge
during winter (Table 1.4). We detected significant differences in seasonal (χ2 = 6.59, d.f.
= 1, p = 0.01) and spatial (χ2 = 176.33, d.f. = 1, p = 0.00) deer availability across the
study area. White-tailed deer comprised 73% and mule deer 27% of prey during summer.
White-tailed deer comprised 68% and mule deer 32% of prey during winter. Annual
availability on The Wedge was 82% white-tailed deer and 18% mule deer, and
availability in Republic was 56% white-tailed deer and 44% mule deer (Table 1.4). Mean
annual prey availability for the entire study area was 70% white-tailed deer and 30%
mule deer.
15
DISCUSSION
Sightability of deer was greater in open habitats (agricultural, open ponderosa
pine, clear cuts) than in dense forested areas (mixed forest), and was greater with
increasing group size, and at shorter distances. We often observed white-tailed deer in
open habitats and at further distances, and observed mule deer in forested habitats closer
to the observer. We expected these differences because white-tailed deer and mule deer
typically use different habitats. Mule deer often prefer forested habitats, which is evident
in a significantly lower sighting distance (perpendicular distance). White-tailed deer
prefer open habitats, such as agricultural areas, and are thus often seen at greater
distances from transects.
Results of the sightability model are consistent with results from the helicopter
survey, suggesting that the corrected ground count is a valid measure of relative deer
abundance. We saw significant variation in prey availability across geographic areas
(The Wedge and Republic) and seasons. However, prey ratios consistently showed a
dominant white-tailed deer population (Table 1.4) across the study area. We expected the
large variations in prey population demographies in the two areas due to differences in
landscape and habitat. Terrain on The Wedge, adjacent to the Columbia River, is
characterized by large swaths of lower elevation agricultural fields, riparian areas, and
deciduous/mixed forest. The Republic are landscape is generally higher in elevation and
dominated by ponderosa pine and Douglas fir forests (Bio/West 1999). We also expected
slight seasonal differences due to a seasonal shift in habitat use. Sightability variance
16
accounted for nearly all of the total variance in the survey because we flew all subunits in
the aerial survey and because we observed deer in different vegetation densities.
Prey availability is often determined by counts (Smith et al. 2004, Honer et al.
2002, Baker et al. 2001, Gil and Pleguezuelos 2000). However, because of the inherent
variability a detection probability (correction factor) should always accompany animal
counts. Of the studies using correction factors (e.g., distance sampling, sightability
models) (Bagchi et al. 2003, Karanth and Sunquist 1995), most are designed such that
transects and units are delineated and randomly chosen at the start of the study, and then
surveyed at regular intervals throughout the study. While these types of static surveys
may be useful to determine prey availability for animals with predictable, stable home
ranges, cougars tend to shift daily movements, monthly home range use, and even entire
home ranges in an unpredictable manner.
Of the 15 cougars that we followed, four of them spent significant time and
occupied substantial area outside the physical boundaries of our 3800 km2 study area.
Additionally, very high annual mortality rates of cougars in our study (Lambert 2003)
required us to frequently shift areas of field work. By surveying deer populations near
daily telemetry locations of live cougars, we were able to determine prey availability in
areas that cougars were actively using. This dynamic approach permitted us to deal more
effectively with the irregular movements of cougars.
Without the flexibility to shift prey count transects so that they reflect cougar
movements, much of the prey availability data collected would be irrelevant. For
example, static surveys, designed within predetermined boundaries, would have omitted
all prey availability in areas of use that fall outside the usual study area boundaries.
17
Conversely, static surveys would have included numerous white-tailed deer along the
highway that were not actually available (within home ranges) to cougars. A dynamic
survey is able to account for such areas, producing more accurate data when determining
prey availability specific to animals.
Biologists are often confronted with two main problems when estimating
population size: observability and sampling. Most animal survey methods do not result
in counts or captures of all animals present in the area. Time and money are frequently
limited, so a particular survey often cannot be applied to the entire area of interest. This
sightability model addresses both of these problems by allowing valid inferences to be
made about the population size from corrected ground counts.
18
LITERATURE CITED
Ackerman, B.R. 1988. Visibility bias of mule deer aerial census procedures in southeast
Idaho. PhD. Diss., University of Idaho., Moscow. 105pp.
Anderson, C.R. and F.G. Lindzey. 1996. Moose sightability model developed from
helicopter surveys. Wildlife Society Bulletin 24:247-259.
Anderson, D.R. 2001. The need to get the basics right in wildlife field studies. Wildlife
Society Bulletin 29:1294-1297.
Anderson, D.R. 2003. Response to Engeman: index values rarely constitute reliable
information. Wildlife Society Bulletin 31: 288-291.
Armeleder, H.M., M.J. Waterhoude, D.G. Keisker, and R.J. Dawson. 1994. Winter
habitat use by mule deer in the central interior of British Columbia. Canadian
Journal of Zoology 72:1721-1725.
Baker, L.A., R.J. Warren, D.R. Diefenbach, and W.E. James. 2001. Prey selection by
reintroduced bobcats (Lynx rufus) on Cumberland Island, Georgia. American
Midland Naturalist 145:80-93.
19
Bagchi, S., S.P. Goyal, and K. Sankar. 2003. Prey abundance and prey selection by tiger
(Panthera tigris) in a semi-arid, dry deciduous forest in western India. Journal of
Zoology, London 260: 285-290.
Bartmann, R. M., G.C. White, L.H. Carpenter, and R.A. Garrott. 1987. Aerial mark-
recapture estimated of confined mule deer in pinyon-juniper woodland. Journal of
Wildlife Management 51:41-46.
Bio/West. 1999. Vegetation mapping on the Okanogan and Colville National Forests
using Landsat Thematic Mapper images. Bio/West, Inc. Logan, UT.
Buckland, S.T., D.R. Anderson, K.P. Burnham, J.L.Laake, D.L. Borchers, and L.
Thomas. 2001. Introduction to Distance Sampling. Oxford University Press. New
York.
Gil, J.M., and J.M. Pleguezuelos. 2001. Prey and prey-size selection by the short-toed
eagle (Circaetusgallicus) during the breeding season in Granada (south-eastern
Spain). Journal of Zoology, London 255:131-137.
Honer, O.P., B. Wachter, M.L. East, and H. Hofer. 2002. The response of spotted
hyaenas to long-term changes in prey populations: functional response and
interspecific kleptoparasitism. Journal of Animal Ecology 71:236-246.
20
Hosmer, D. W., and S. Lemeshow. 1989. Applied logistic regression: John Wiley &
Sons, New York.
Karanth, K.U. and M.E. Sunquist. 1995. Prey selection by tiger, leopard and dhole in
tropical forests. Journal of Animal Ecology 64:439-450.
Kleinbaum, D.G, L.L. Kupper, and L.E. Chambless. 1982. Logistic regression analysis of
epidemiologic data: theory and practice. Communications in Statististics: Theory
and Methods 11:485-547.
Krebs, C.J. 1999. Ecological methodology, Second edition. Harper and Row, Publ., New
York.
Lambert, C.M. 2003. Cougar population dynamics and viability in the Pacific Northwest.
M.S. Thesis. Washington State University. 38pp.
Manly, B., L. McDonald, and D. Thomas. 1993. Resource selection by animals: statistical
design and analysis for field studies. Chapman & Hall, New York.
Ott, R.L., and M. Longnecker. 2001. An introduction to statistical methods and data
analysis. 5th edition. Duxbury. Pacific Grove, PA.
21
Pauley, G.R., J.M. Peek, and P. Zager. 1993. Predicting white-tailed deer habitat use in
northern Idaho. Journal of Wildlife Mangement 57:904-913.
Otten, R.M., J.B. Haufler, S.R. Winterstien, and L.C. Bender. 1993. An aerial census
procedure for elk in Michigan. Wildlife Society Bulletin 21:73-80.
Robinson, H.S., R.B. Wielgus, and J.C. Gwilliam. 2002. Cougar predation and population
growth of sympatric mule deer and white-tailed deer. Canadian Journal of
Zoology 80:777-790.
Samuel, M.D., E.O. Garton, M.W. Schlegel, and R.G. Carson. 1987. Visibility bias
during aerial surveys of elk in northcentral Idaho. Journal of wildlife Management
51:622-630.
Smith, D.W., T.D. Drummer, K.M. Murphy, D.S. Guernsey, and S.B. Evans. 2004.
Winter prey selection and estimation of wolf kill rates in Yellowstone National
Park. Journal of Wildlife Management 68: 153-165.
Steinhorst, R.K. and M.D. Samuel. 1989. Sightability adjustment methods for aerial
surveys of wildlife populations. Biometrics 45:415-425.
Thomas, L., J.L. Laake, S. Strindberg, F.F.C. Marques, S.T. Buckland, D.L. Borchers,
D.R. Anderson, K.P. Burnham, S.L. Hedley, and J.H. Pollard. 2002. Distance 4.0.
22
Release 11. Research Unit for Wildlife Population Assessment, University of St.
Andrews, UK. http://www.ruwpa.st-and.ac.uk/distance/.
Unsworth, J. W., F. A. Leban, E. O. Garton, D. J. Leptich, and P. Zager. 1999. Aerial
survey: user’s manual. Electronic Edition. Idaho Department of Fish & Game,
Boise, Idaho, USA.
Williams, B.K., J.D. Nichols, and M.J. Conroy. 2002. Analysis and management of
animal populations: modeling, estimation, and decision making. Academic Press,
San Diego, CA.
23
Table 1.1. Process of model selection for a logistic regression analysis of deer
sightability in northeastern Washington, 2002-2004.
Univariate Results Log-likelihood χ2 df p-value Improvement χ2 df p-value Rho-rsqb
Habitat 35.215 4 0.000 0.231
Habitat, Distance 59.493 5 0.000 24.278 1 0.000 0.391
Habitat, Dist., Grp Sizea 72.391 6 0.000 12.898 1 0.003 0.476 aModel with the best fit
bMcFaddens’ Rho-square
24
Table 1.2. Coefficients, standard errors, probabilitites of significance (t-ratio), and odds
ratios for group size, distance and habitat parameters used in a logistic regression model
for deer sightability in northeastern Washington during fall 2003.
Parameter β SE P Odds Ratioa
β1 Group Size 0.682 0.222 0.002 0.505
β2 Distance -0.044 0.011 0.000 1.045
β3 Habitat (Clear Cut) 0.463 0.526 0.379 0.629
β3 Habitat (Forest) -3.291 0.682 0.000 26.864
β3 Habitat (Agricultural) 2.498 0.738 0.001 0.082
β3 Habitat (P Pine) 0.330 0.500 0.510 0.719
Intercept 1.868 0.788 0.018
aOdds ratio = Exp (β); the factor by which the odds that a deer will be sighted change for
every unit increase in the independent variable.
25
Table 1.3. Raw (obs) and corrected (cor) numbers of deer observed in each of four habitat types from ground surveys in 2003 for
northeastern WA.
Habitat TypesClear Cut Forest Agricultural Ponderosa Pine Totals
Deer Species Obs Cor Obs Cor Obs Cor Obs Cor Obs Cor
Mule DeerIndividual 17 18 155 431 152 159 31 31 355 639
Group 10 10 96 96 31 31 13 13 150 150
White-tailed DeerIndividual 25 25 242 981 520 550 56 56 843 1612
Group 16 16 135 135 138 138 28 28 317 317
Total DeerIndividual 42 43 397 1412 672 709 87 87 1198 2251
Group 26 26 231 231 169 169 41 41 467 467
26
Table 1.4. Estimated total (N) and relative (%) numbers of white-tailed deer and mule
deer from corrected ground and aerial surveys in 2002-2004 in northeastern WA.
White-tailed Deer Mule Deer
N % N %
Ground Survey Annual Study Area 1612 (71.6) 639 (28.4)
Wedge 1130 (81.6) 255 (18.4)
Republic 482 (55.7) 384 (44.3)
Summer Study Area 1139 (73.3) 416 (26.8)
Wedge 783 (85.) 138 (15.)
Republic 356 (56.1) 278 (43.9)
Winter Study Area 473 (68.) 223 (32.)
Wedge 400 (77.8) 114 (22.2)
Republic 73 (51.8) 109 (48.2)
Aerial Survey Winter Wedge 686 (77.8) 196 (22.2)
27
CHAPTER 2
PREY SELECTION AND KILL RATES OF COUGARS
IN NORTHEASTERN WASHINGTON
ABSTRACT
We investigated prey selection of cougars in northeastern Washington during
2002-2004, where sympatric white-tailed deer and mule deer are the primary and
secondary prey species, respectively. We tested two competing hypotheses of prey
selection, the “prey switching” hypothesis, and the “apparent competition” hypothesis. .
To estimate use of prey by cougars, we examined 60 cougar kills. White-tailed deer
comprised 60% of the kills (mule deer comprised 40%), a proportion larger than the
study area’s prey population (70% white-tailed deer vs. 30% mule deer). Cougars
selected for mule deer across the entire study area. We also detected strong seasonal
changes in prey selection, with cougars strongly selecting for mule deer in summer, but
not during winter. Mean annual kill rate was 6.68 days per deer killed. Kill rates did not
differ between seasons or deer species. Habitat characteristics of kill sites did not differ
significantly between white-tailed deer and mule deer kills. These findings are consistent
with the apparent competition hypothesis and suggest that the current decline in mule
deer numbers in northeastern Washington is caused by an abundant invading primary
prey species (white-tailed deer) and a related increase in predation on the secondary prey
species (mule deer) during summer months.
28
INTRODUCTION
Within the last ten years, a major change in the population structure of deer in
Western North America has taken place. Native mule deer (Odocoileus hemionus)
populations have sharply declined, while non-native white-tailed deer (Odocoileus
virginianus) populations have increased (Gill 1999). Recent work on sympatric deer in
California (Bleich and Taylor, 1998) and south-central British Columbia (Robinson et al.
2002) showed that cougar predation was a major mortality factor for both deer species.
Robinson et al. (2002) found that increasing sympatric white-tailed deer (primary prey)
were more numerous than declining mule deer (secondary prey), but that intrinsic growth
rates (birth rates) were not different between the two species. The study also found that
per capita predation rates by cougars were greater for mule deer than for white-tailed
deer, especially during summer, and that predation rates increased with increasing prey
density for white-tailed deer, but increased with decreasing prey density for mule deer.
Two hypotheses could explain these differences in predation, including the
“apparent competition” hypothesis, (Holt 1977), and the “prey switching” hypothesis
(Holling 1961). The apparent competition hypothesis predicts that a primary prey species
can increase predator numbers; thereby having a negative effect on a secondary prey
through the commonly shared predator. The secondary prey can decline through, 1)
lower reproductive rates if predation rates are proportionate to abundance, or 2)
disproportionate predation rates (prey selection) if reproductive rates are similar. Prey
switching occurs when the focus of a predator is switched from one prey type to another
after the “alternate” prey species increases beyond some threshold density. The
29
secondary prey can decline because of disproportionate predation caused by a shift in the
predator’s search image and/or habitat use.
If apparent competition is occurring in a predator-prey community, we would
expect to see cougars following their primary prey and residing in its habitat (white-tailed
deer), while killing (selecting) the secondary prey, mule deer, at a higher rate in the same
habitats. Therefore, cougars would select for, or disproportionately kill, mule deer
simply because they have greater success killing them than white-tailed deer. The kill
rates should be equal to or higher than that of white-tailed deer despite greater body mass
in mule deer. If prey switching is occurring, we would expect to see a shift in habitat use
by the predator, as predators change their search image to seek out the more abundant
secondary prey species. In this case, kill rate (days/kill) should also be lower when
cougars prey on mule deer because of their greater body mass (Silva and Downing 1995).
Seasonal abundance of mule deer should surpass white-tailed deer concurrent with the
prey shift. Habitats at mule deer kill sites should be different than white-tailed deer kill
sites; indicating that cougars purposefully shift to and select for mule deer in their
associated habitats.
In this paper, we test Robinson’s (2002) prediction that cougars select for mule
deer over sympatric white-tailed deer. We also test the apparent competition and prey
switching hypotheses, if such selection occurs. Specifically, our objectives were to 1)
determine the relative proportions of white-tailed deer and mule deer available to
cougars, 2) determine the relative proportions of cougar-killed white-tailed deer and mule
deer, 3) determine the kill rate (kills/cougar/unit time) for white-tailed deer and mule
deer, and 4) compare habitat characteristics at white-tailed deer and mule deer kill sites.
30
STUDY AREA
The study area covers approximately 3,800 km2 in northeastern Washington.
Boundaries extend from the Okanogan/Ferry County line west to the Columbia River,
and from the Canadian/US border south to the Colville Indian Reservation. The study
area lies entirely within the Okanogan Highlands physiographic province, composed of
glacially subdued mountainous terrain, with elevations ranging from 400 m in the
Columbia River valley to 2130 m at the top of the Kettle Crest. Forest overstory species
include Douglas-fir (Pseudotsuga menziesii), western hemlock (Tsuga heterophylla),
western red cedar (Thuja plicata), and subalpine fir (Abies lasiocarpa). Inland maritime
conditions characterize the climate, with mean temperatures ranging from –6 °C in
January to 21 °C in July, and annual precipitation of 46 cm. Snowfall averages 136 cm
during a 6-month period from mid-November to mid-April.
Field efforts were concentrated in two separate areas separated by the Kettle
River. “The Wedge” study area lies to the east of the Kettle River, and the “Republic”
area lies to the west of the River and the Kettle Crest Mountain Range. We conducted
analyses on each replicate study area separately.
Robinson et al. (2002) found that white-tailed deer (Odocoileus virginianus) were
the most abundant ungulate in a nearby study area, followed by mule deer (Odocoileus
hemionus). Since climate and physiography create seasonally migratory deer
populations, both white-tailed deer and mule deer congregate on winter ranges between
December and April. Deer winter ranges are generally on south to west-facing gentle
slopes in timber stands with higher canopy closure, providing wind and snow breaks
31
(Pauley et al. 1993, Armeleder et al. 1994). Across the study area, higher elevation
winter ranges (i.e., 900-1200 m) are almost exclusively occupied by mule deer, whereas
lower elevation ranges (i.e., ≤ 900 m) are predominantly used by white-tailed deer.
During summer white-tailed deer move up in elevation and intermix with mule deer.
32
METHODS
Cougar Kills
We located cougars by observing cougar tracks from snowmobiles, trucks, and on
foot. Trained hounds treed cougars, which were then immobilized with a 2ml mixture of
Ketamine hydrochloride (3 mg/kg) and Medetomidine hydrochloride (.08 mg/kg)
delivered via brown charged aluminum darts from a Cap Chur rifle (Cap-Chur, Inc.,
Powder Springs, GA). We used a drop net at the base of the tree to reduce any injuries
from a fall. If a cougar remained in a tree, it was lowered to the ground using a rope
secured around its chest or shoulder.
While immobilized, we classified each cougar into one of three age classes (kitten
< 1yr, subadult 1-2 yrs, adult > 2yrs) based on gum regression (Shaw 1987). We fitted
the cougar with two numbered ear tags and a mortality-sensitive VHF radio-collar. For
details on capturing and collaring see Lambert (2003).
We located cougars approximately once per week using fixed wing aircraft. Two
H antennas mounted underneath each wing allowed us to accurately determine cougar
positions, which we marked with a Garmin (Garmin International, Inc., Olathe, KS)
Global Positioning System (GPS) unit. We also kept track of cougar movements using
ground telemetry and/or snow tracking several times per week on cougars throughout the
study. Individual cougars were monitored over 21-day predation sequences (Murphy
1998, Nowak 1999). Daily locations were determined by plotting three or more
converging bearings taken in the field. Error polygons were established in LOAS
(Location of a Signal) telemetry triangulation software (Ecological Software Solutions,
33
Sacramento, CA), and plotted on 1:24,000 United States Geological Survey topographic
maps.
We discovered cougar kills by visiting sites that cougars occupied the previous
day, and searched for a kill with the help of hounds. Searched locations were typically
100 m x 100 m (1 ha) in size. When a kill was located, we determined whether cougars
were responsible based on the proximity to cougar locations and field sign (e.g., cougar
bed site, cougar tracks or scat, and signs that the carcass had been covered or cached)
(Shaw 1987). We determined deer species by examining the metatarsal gland on the
outside of the lower hind leg. White-tailed deer have a relatively small, white metatarsal
gland (30 mm long), whereas mule deer have a larger (50-150 mm long), darker
metatarsal gland (Verts and Carraway 1998).
Time since the kill was calculated by recording the first day that the cougar was
located at the site, and by noting the condition of blood and exposed muscle on the
carcass, presence of maggots on the carcass, and knowledge of recent weather conditions
that may affect the condition of the kill. We then assigned a kill date to the carcass, and
calculated a predation sequence (interval, in days, between two consecutive kills) for
each cougar.
Prey Availability
We determined relative prey availability for collared cougars using a dynamic
sampling technique that followed cougar movements throughout their home ranges.
While monitoring collared cougars, we recorded all live white-tailed deer and mule deer
encountered on foot or in a vehicle. To ensure that deer were truly available to cougars,
34
we included only those deer within the cougar home range. On occasions when transects
(roads, trails) were covered more than once per day, we recorded deer sightings only one
time per day.
Each time a group or a single deer was observed, the date, time, Universal
Transverse Mercator (UTM) coordinate, species, number of animals, sex, age, habitat,
and the straight-line distance from our position were recorded. We classified deer as
male or female, and adult, yearling, or fawn. Habitat types were classified as open
ponderosa pine, dense mixed forest, clear-cut, or agricultural. To incorporate an estimate
of detection probability to our deer count, we developed a sightability model using life-
sized deer decoys as “marked deer” (see chpt.1). We used logistic regression (LR)
analyses (Proc Logistic, SAS Institute, Cary N.C.) to test for effects of group size,
distance, and habitat on deer sightability, then used that LR model on deer sightability to
calculate corrected relative numbers of live white-tailed deer and mule deer on survey
transects.
We also conducted a late winter helicopter survey to estimate relative prey
availability on one part of the study area (The Wedge) to compare against the ground
count. Before the survey, we identified 20 subunits distinguishable from the air by roads,
drainages, and topographic features. Size of subunits ranged from about 4 km2 to 17 km2
and each required less than one hour to survey. The data collected from our ground
counts, along with knowledge of the district biologists helped us to define and stratify
subunits as low or high white-tailed deer and mule deer density (Robinson et al. 2002).
In February 2004, we surveyed all 20 subunits following methods of Unworth et
al (1999)to determine relative white-tailed deer and mule deer availability. The same
35
primary observer and pilot were used throughout the survey, while secondary staff varied
from day to day. Point estimates, variances, and confidence intervals were calculated
based on formulas in the Mule Deer Spring model in program AERIAL SURVEY
version 1.0 (Unsworth et al. 1999).
We compared the proportion of white-tailed deer and mule deer from the
corrected winter ground survey on The Wedge to that from the aerial survey. We also
compared seasonal vs. annual proportions, and proportions of deer species in the entire
study area vs. the two sub-areas (The Wedge, Republic). We used the chi square test of
homogeneity for all comparisons.
Prey Selection and Kill Rates
Because of small sample sizes of kills for individual cougars, we tested if 2nd
order landscape availabilities (The Wedge and Republic) were the same as 3rd order home
range availabilities (Johnson 1980). Using 2nd order availabilities would allow us to
include kills of animals with unspecified home ranges (< 32 radiolocations). We
compared relative availability of white-tailed deer and mule deer within the 95% adaptive
kernel home range for each cougar to relative availabilities on a broader, composite home
range, landscape scale (The Wedge and Republic) using the t-test. If home range and
landscape availabilities were not different, we used landscape availabilities and kills of
all cougars to test for prey selection. We calculated a 95% adaptive kernel home range
for each cougar with greater than 32 locations using the animal movement extension for
ArcView (ESRI, Redlands, California, U.S.A.). Thirty-two locations were found to be
adequate to describe home ranges by Logan and Sweanor (2001), additionally 30
36
locations is sufficient to approximate the standard normal distribution (Zar 1984). We
experienced small sample sizes of kills for individual cougars because of very high
cougar mortality across the study area; very few cougars lived for > 1 year (Lambert
2003).
We then tested for prey selection by comparing observed and expected numbers
of white-tailed deer and mule deer kills for each individual cougar using the chi-square
goodness of fit test. Expected kills were calculated by multiplying the total number of
kills by the relative availability of each species in The Wedge and Republic. Because of
small sample sizes of kills for each animal, we tested if kills could be pooled across
cougars using the heterogeneity chi-square test (Zar 1984). If the heterogeneity chi-
square test was not significant, we totaled all observed and expected values for deer in
each study area (The Wedge and Republic) and performed a chi-square goodness of fit
test on the totals. We performed this analysis on The Wedge and Republic separately. If
pooling was justified for each area, we then performed the heterogeneity chi-square test
on the study area as a whole. Once again, if differences were not significant, we pooled
across both areas to increase sample sizes of kills.
To analyze seasonal selection, we used summer (May 1 through October 31) and
winter (November 1 through April 30) observed and expected values. We used seasonal
species-specific availabilities and deer kills from The Wedge and Republic to calculate
expected values. A heterogeneity chi-square test was used to determine if seasonal
selection could be pooled across cougars. If chi-square test of heterogeneity yielded no
differences in selection of prey within the entire study area geographic area, we pooled
seasonal selection across the entire study area.
37
For each of the preceding selection analyses, we calculated mean use/availability,
or selection ratios (Manly et al. 1993), and tested for differences in mean selection ratios
for mule deer and white-tailed deer with a t-test. We used a one-tailed test in selection
ratio analyses because we hypothesized that cougars select for mule deer over white-
tailed deer.
To estimate kill rate of cougars from predation sequences, we calculated the
number of days between two consecutive deer kills (inter-kill interval). We pooled all
inter-kill intervals for a mean annual and seasonal kill rate. Summer intervals occurred
from 1 May through 31 October, and winter intervals occurred from 1 November through
30 April. To determine if kill rate differed between seasons, we used a t-test. We also
used a t-test to compare kill rate following white-tailed deer kills to kill rate following
mule deer kills.
Habitat Characteristics at Kill Sites
To determine whether mule and white-tailed deer were killed in different habitats,
we recorded 9 habitat variables within a 25 m radius plot from the site center of each kill
site. Variables included elevation, physiography (ridge crest, stream valley, cliff/rock
bench, open slope, or forested slope), habitat type (mixed conifer, mixed forest,
ponderosa pine, shrub steppe, riparian, clear-cut, or agricultural), tree species, slope,
aspect, snow depth, canopy density (measured with a densitometer), shrub density
(measured visually by estimating percent cover of 5.64 m radius circle surrounding kill).
We used a chi-squared test of homogeneity to test for differences in categorical variables
(physiography, habitat type, and aspect) between white-tailed deer kill sites and mule
38
deer kill sites. A t-test was used to compare continuous variables (elevation, slope, snow
depth, canopy density, and shrub density) between kill sites of mule deer and white-tailed
deer. We tested for seasonal differences in kill sites (summer vs. winter) using identical
techniques.
39
RESULTS
Cougar Kills
During 2002 through 2004, we monitored 14 (12 females and 2 males) of 19
collared cougars to determine prey selection and kill rate (5 cougars were inaccessible or
died shortly after capture). We completed 27, 21-day monitoring sequences. Fourteen
additional sequences were incomplete, as the animal moved into an inaccessible location,
across the US/Canadian border, out of the study area.
From May 2002 to March 2004, we found 60 cougar-killed deer. Cougars killed
more white-tailed deer during both winter (χ2 = 5.14, d.f. = 1, p = 0.02) and summer
seasons (χ2 = 3.5, d.f. = 1, p = 0.06). We identified 35 white-tailed deer kills (58%), 23
mule deer kills (38%), but were unable to differentiate species of 2 kills (4%). During
winter, white-tailed deer comprised 65% of kills and mule deer comprised 35% of kills.
During summer, white-tailed deer comprised 57% of kills and mule deer comprised 38%
of kills (5% of kills were unidentified).
Prey Availability
Raw counts of deer available to collared cougars indicated that white-tailed deer
were more abundant than mule deer, with more groups (317 vs. 150) and total individuals
(843 vs. 355). The uncorrected relative deer population across the entire study area (The
Wedge and Republic) was 70% white-tailed deer and 30% mule deer. Our sightability
model (see Chapter 1) increased the estimated number of white-tailed deer from 843 to
1612 (+ 191%), and mule deer from 355 to 639 (+ 180%). The corrected relative
40
availability across the study area was 72% white-tailed deer and 28% mule deer. The
February 2004 aerial survey on The Wedge also indicated that white-tailed deer
comprised the bulk of the deer population (χ2 = 2997, d.f. = 1, p < 0.001), with an
estimated population size of 1,384 ± 221 white-tailed deer (80%) and 354 ± 83 mule deer
(20%).
We detected significant differences in seasonal (χ2 = 6.59, d.f. = 1, 0.01 < p <
0.025) and spatial (χ2 = 176.33, d.f. = 1, p < 0.001) deer availability across the study area
(see chpt.1). White-tailed deer comprised 68% and mule deer 32% in winter. White-
tailed deer comprised 73% and mule deer comprised 27% in summer. Annual
availability on The Wedge was 82% white-tailed deer and 18% mule deer, and
availability in Republic was 56% white-tailed deer and 44% mule deer (Table 2.1). Chi
square results yielded no difference (χ2 = 0.0002, d.f. = 1, p > 0.10) between the ground
survey and helicopter survey on The Wedge during winter. Percentages of available deer
appeared higher than the percentages of kills for white-tailed deer (70% available vs.
60% killed) and lower than the percentages of kills for mule deer (30% available vs. 40%
killed). At no time or location did mule deer abundance equal or exceed white-tailed deer
abundance.
Prey Selection and Kill Rates
Mean prey availability within home ranges was not different from landscape
availabilities on The Wedge (t = -0.34, p = 0.75) and Republic (t = 0.49, p = 0.67),
allowing landscape availabilities to be used to test for prey selection. Mean home range
availabilities on The Wedge were 0.83 (± 0.09 SD) for white-tailed deer and 0.17 (± 0.09
41
SD) for mule deer compared to 0.82 and 0.18 for landscape availability across The
Wedge. Mean home range availabilities in Republic were 0.58 (± 0.08 SD) for white-
tailed deer and 0.42 (± 0.08 SD) for mule deer compared to 0.56 and 0.44 for landscape
availability across Republic.
At the 3rd order (home range) of selection, only 3 of 13 individual cougars
selected for mule deer (Table 2.1). However, chi-square tests of heterogeneity yielded no
differences among selection, allowing cougars to be pooled within areas (Table 2.1).
Pooled chi square goodness of fit values yielded significant 2nd order (landscape)
selection for mule deer on The Wedge (χ2 = 2.82, d.f. = 1, p = 0.09), but not in Republic
(χ2 = 1.99, d.f. = 1, p = 0.16) (Table 2.1). Mule deer selection ratios across the study area
were not different from white-tailed deer ratios on The Wedge (t = -1.37, d.f. = 14, p =
0.10) and Republic (t = -1.09, d.f. = 10, p = 0.15) (Table 2.1). However, chi square tests
of heterogeneity yielded no differences among The Wedge and Republic (χ2 = 11.28, d.f.
= 13, p = 0.59), allowing cougars to be pooled across areas. Pooled chi square goodness
of fit tests indicated that cougars selected for mule deer across the entire study area (χ2 =
4.42, d.f. = 1, p = 0.04). The mule deer selection ratio (1.53) for the entire study area was
also higher than the white-tailed deer ratio (0.82) (t = -1.75, d.f. = 26, p = 0.05).
Cougars strongly selected for mule deer during summer (χ2 = 4.28, d.f. = 1, p =
0.04), but not during winter (χ2 = 0.04, d.f. = 1, p = 0.84) (Table 2.2). Selection ratios
also showed significant selection for mule deer (1.441 vs. 0.829) over white-tailed deer,
t= -1.51, d.f. = 22, p = 0.07) in summer, but not during winter (1.043 vs. 1.028) (t = 0.04,
d.f. = 16, p = 0.49).
42
Mean annual kill rate of cougars was 6.68 days per deer kill (SD = 3.12; range =
2.0-14.0, n = 22 sequences). Kill rates did not differ between seasons (6.56 days/kill for
summer and 7.0 days/kill for winter, t = -0.29, df = 20, p = 0.78) or deer species (7.00
days/kill for white-tailed deer and 6.14 days/kill for mule deer, t = 0.58, df = 19, p =
0.58).
Habitat Characteristics at Kill Sites
We assessed habitat variables at 55 kill sites (30 white-tailed deer and 25 mule
deer) (Tables 2.3 & 2.4). All chi-square tests showed no difference (χ2 = 0.18 – 0.85, d.f.
= 1-3, p > 0.05) between white-tailed deer kill sites and mule deer kills sites. T-tests (t =
0.16 - 0.95, d.f. = 23-53, p > 0.05) also showed no differences in mule deer vs. white-
tailed deer kill sites for the entire study area. We found a seasonal difference in habitat
type (χ2 = 6.63, d.f. = 2, p = 0.04) at kill sites, however no other habitat characteristics
showed seasonal differences. When broken into geographic areas, mule deer kills were
located in higher elevations than white-tailed deer kills in The Wedge during summer (t =
1.91, d.f. = 31, p = 0.07). Small sample sizes prevented me from analyzing kill sites in
Republic.
43
DISCUSSION
Across the study area and within The Wedge, cougars selected for mule deer over
white-tailed deer during the year. When examined seasonally, cougars strongly selected
for mule deer during the summer but not during the winter, and in no season or location
did they select for white-tailed deer. The annual kill rate of 7 days for cougars falls
within the range of 7 to 11 days reported by other investigators (Hornocker 1970, Beier et
al. 1995, and Murphy 1998). The interval may be at the low end because 15 of the 22
intervals were from female cougars with kittens, which typically show a higher kill rate
than single adults (Murphy 1998). Only 2 intervals were from a male cougar (8 and 11
days). We found no differences in habitat characteristics between mule deer and white-
tailed deer kill sites.
Our results indicate that cougars select for mule deer on a seasonal basis. White-
tailed deer comprised the primary prey during both seasons, but disproportionate
predation of mule deer occurred during the summer as cougars followed white-tailed deer
into mule deer range during that season. Furthermore, our results on kill rate and habitat
use are inconsistent with the “prey switching” hypothesis. Mule deer availability never
equaled or exceeded white-tailed deer availability. Cougars showed no shift in habitat
use at mule deer and white-tailed deer kill sites, and there were no differences in kill rates
for white-tailed deer and mule deer; suggesting that cougars did not switch their search
image to seek out more numerous, larger mule deer during the summer. These findings
support the apparent competition hypothesis. Total numbers of cougar kills indicate that
white-tailed deer are the primary prey of cougars (Tables 2.1 & 2.2). However, to fully
44
understand predator-prey relationships, both use and availability must be taken into
consideration (Johnson 1980). Cougars did not select for white-tailed deer, but selected
for mule deer during the summer. It is evident that cougars, while primarily subsisting on
white-tailed deer, are disproportionately affecting the mule deer population, as suggested
by Robinson et al. (2002).
The similar kill rates between mule deer and white-tailed deer also suggest that
cougars select for, or disproportionately kill, mule deer during the summer because they
are easier to kill. The larger body mass of mule deer should result in a greater inter-kill
interval, but this was not the case. Lingle (2002) suggested different anti-predator
behaviors and escape mechanisms for white-tailed deer and mule deer as reasons for
predator effectiveness. As a first line of defense, mule deer remain in high and rugged
habitats as much as possible to minimize their exposure to predators and dissuade attacks.
However, once encountered, mule deer are slower, and less able to avoid attack and
capture. This appears to be the case in our study area.
From the seasonal selection that we found and the elevational shift of kill sites
from lower to higher in summer, it is apparent that during winter, cougars occupy lower
elevations and gentle slopes typical of white-tailed deer winter ranges (Pauley et al. 1993,
Armeleder et al. 1994); and in the summer shift their home range use, following the
elevational migration of white-tailed deer. When cougars move into higher terrain, they
are more likely to overlap areas used by mule deer (rugged terrain, steep slopes, avg.
summer elevation = 1800m) (Pauley et al. 1993, Armeleder et al. 1994). As a result,
incidental encounters between mule deer and cougars increase. Katnick (2002) found
similar relationships with cougar predation of white-tailed deer and mountain caribou.
45
His results showed annual landscape scale selection for white-tailed deer, and seasonal
(summer) selection for caribou. He attributed this to a shift in elevation by deer, causing
a high amount of spatial overlap between cougars and caribou during the summer
months. Results of Robinson et al. (2002) also suggest this seasonal pattern of predation.
In their area, the greatest difference in cougar predation rates between white-tailed deer
and mule deer occurred during the summer.
Robinson et al. (2002) first suggested that the patterns of cougar predation on
white-tailed deer and mule deer population growth rates fit the apparent competition
theory. His observed mule deer decline, although directly attributed to cougars, was
ultimately caused by an abundance of invading primary prey (white-tailed deer). As
white-tailed deer numbers are increasing, mule deer have become secondary prey species,
and are now at risk of depensatory predation. We urge other researchers to test for
seasonal prey selection and apparent competition in systems with an invading, non-native
primary prey and a declining secondary prey species.
46
LITERATURE CITED
Armeleder, H.M., M.J. Waterhouse, D.G. Keisker, and R.J. Dawson. 1994. Winter
habitat use by mule deer in the central interior of British Columbia. Canadian
Journal of Zoology 72:1721-1725.
Beier, P., D. Choate, and R.H. Barrett. 1995. Movement patterns of mountain lions
during different behaviors. Journal of Mammalogy 76:1056-1070.
Berger, J., and J.D. Wehausen. 1991. Consequences of mammalian predator-prey
disequilibrium in the Great Basin Desert. Conservation Biology 5:244-248.
Bleich, V.C., and T.J.Taylor. 1998. Survivorship and cause specific mortality in five
populations of mule deer. Great Basin Naturalist 58:265-272.
Gill, R.B. 1999. Declining mule deer populations in Colorado: reasons and responses.
Colorado Division of Wildlife, Denver, Colorado.
Holling, C.S. 1961. Principles of insect predation. Annual Review of Entomology 6:163-
182.
Holt, R.D. 1977. Predation, apparent competition, and the structure of prey communities.
Theoretical Population Biology 12:197-229.
47
Hornocker, M.G. 1970. An analysis of mountain lion predation upon mule deer and elk
in the Idaho Primitive Area. Wildlife Monographs 21:3-39..
Johnson, D.H. 1980. The comparison of usage and availability measurements for
evaluating resources preference. Ecology 61:65-71.
Katnick, D. D. 2002. Predation and habitat ecology of mountain lions (Puma concolor) in
the southern Selkirk Mountains. Dissertation. Washington State University.
217pp.
Lambert, C.M. 2003. Dynamics and viability of a cougar population in the Pacific
Northwest. M.S. Thesis. Washington State University. 35pp.
Lingle, S. 2002. Coyote predation and habitat segregation of white-tailed deer and mule
deer. Ecology 83:2037-2048.
Logan, K.A. and L.L. Sweanor. 2001. Desert Puma. Island Press. USA.
Manly, B., L. McDonald, and D. Thomas. 1993. Resource selection by animals: statistical
design and analysis for field studies. Chapman & Hall, New York.
48
Murphy, K.M. 1998. The ecology of the cougar in the Northern Yellowstone Ecosystem:
interactions with prey, bears, and humans. Dissertation. University of Idaho.
147pp.
Nowak, C.M. 1999. Predation rates and foraging ecology of adult female mountain lions
in northeastern Oregon. M.S. Thesis. Washington State University. 76pp.
Pauley, G.R., J.M. Peek, and P. Zager. 1993. Predicting white-tailed deer habitat use in
northern Idaho. Journal of Wildlife Management 57:904-913.
Robinson, H.S., R.B. Weilgus, and J.C. Gwilliam. 2002. Cougar predation and population
growth of sympatric mule deer and whit-tailed deer. Canadian Journal of Zoology
80:556-568.
Shaw, H.G. 1987. Mountain lion field guide. Special Report Number 9. Arizona Game
and Fish Department. Third Edition. 47 pp.
Silva, M. and J.A. Downing. 1995. The CRC handbook of mammalian body mass. CRC
Press. Boca Raton, Florida. USA.
Unsworth, J. W., F. A. Leban, E. O. Garton, D. J. Leptich, and P. Zager. 1999. Aerial
Survey: User’s Manual. Electronic Edition. Idaho Department of Fish & Game,
Boise, Idaho, USA.
49
Verts, B.J., and L.N. Carraway. 1998. Land mammals of Oregon. University of
California Press, Los Angeles, CA.
Wehausen, J.D. 1996. Effects of mountain lion predation on bighorn sheep in the Sierra
Nevada and Granite Mountains of California. Wildlife Society Bulletin 24:471-
479.
Zar, J.H. 1984. Biostatistical Analysis, 2nd ed. Prentice-Hall, Inc. Englewood Cliffs, NJ.
50
Table 2.1. Chi-square test results and selection ratios for 2nd order cougar prey selection
in northeastern Washington, during 2002-2004. Prey availability within study areas,
observed deer kills, and expected deer kills are given for white-tailed deer (WT) and
mule deer (MD). Omnibus statistics show results of pooling across areas and cougars.
Availability Observed Expected Selection Ratios
Cougar WT MD WT MD WT MD χ2 p WT MD
Wedge 102 0.82 0.18 3 0 2.45 0.55 0.68 0.41 1.23 0.00
223 0.82 0.18 5 3 6.53 1.47 1.94 0.16 0.77 2.04
293 0.82 0.18 7 1 6.53 1.47 0.19 0.67 1.07 0.68
402 0.82 0.18 0 1 0.82 0.18 4.56 0.03 0.00 5.56
423 0.82 0.18 3 3 4.90 1.10 3.99 0.05 0.61 2.72
662 0.82 0.18 5 2 5.71 1.29 0.48 0.49 0.88 1.55
473 0.82 0.18 1 0 0.82 0.18 0.23 0.63 1.23 0.00
683 0.82 0.18 3 1 3.26 0.74 0.12 0.73 0.92 1.36
Total χ2 12.17 0.14
Pooled χ2 27 11 31.01 6.99 2.82 0.09 0.84 1.74
Heterogeneity χ2 9.35 0.23
Republic 145 0.56 0.44 0 1 0.56 0.44 1.26 0.26 0.00 2.26
154 0.56 0.44 1 6 3.90 3.10 4.85 0.03 0.26 1.93
191 0.56 0.44 1 0 0.56 0.44 0.80 0.37 1.80 0.00
261 0.56 0.44 4 2 3.34 2.66 0.29 0.59 1.20 0.75
341 0.56 0.44 1 1 1.11 0.89 0.03 0.87 0.90 1.13
593 0.56 0.44 1 2 1.67 1.33 0.61 0.44 0.60 1.50
Total χ2 6.58 0.36
Pooled χ2 8 12 11.13 8.87 1.99 0.16 0.79 1.26
Heterogeneity χ2 4.59 0.47
Omnibus χ2 13.99 0.45
Pooled χ2 35 23 42.14 15.86 4.42 0.04 0.82 1.53
Heterogeneity χ2 11.28 0.59
51
Table 2.2. Chi-square test results and selection ratios for seasonal 2nd order cougar prey
selection in northeastern Washington, during 2002-2004. Prey availability within cougar
home ranges, observed deer kills, and expected deer kills are given for white-tailed deer
and mule deer.
Availability Observed Expected Selection Ratios
Cougar WT MD WT MD WT MD χ2 p WT MDSummer Wedge 102 0.85 0.15 3 0 2.55 0.45 0.53 0.47 1.18 0.00
223 0.85 0.15 3 2 4.25 0.75 2.46 0.12 0.71 2.67
293 0.85 0.15 5 1 5.10 0.90 0.01 0.91 0.98 1.11
423 0.85 0.15 2 2 3.40 0.60 3.85 0.05 0.59 3.34
473 0.85 0.15 1 0 0.85 0.15 0.18 0.67 1.18 0.00
662 0.85 0.15 1 1 1.70 0.30 1.93 0.17 0.59 3.34
683 0.85 0.15 2 0 1.70 0.30 0.35 0.55 1.18 0.00
Republic 154 0.56 0.44 0 4 2.25 1.75 5.12 0.02 0.00 2.28
191 0.56 0.44 1 0 0.56 0.44 0.78 0.38 1.78 0.00
261 0.56 0.44 2 2 2.25 1.75 0.06 0.80 0.89 1.14
341 0.56 0.44 1 1 1.12 0.88 0.03 0.86 0.89 1.14
593 0.56 0.44 0 1 0.56 0.44 1.28 0.26 0.00 2.28
Total χ2 16.58 0.17
Pooled χ2 21 14 26.29 8.71 4.28 0.04 0.829 1.441
Heterogeneity χ2 12.30 0.34Winter Wedge 223 0.78 0.22 2 0 1.56 0.44 0.57 0.45 1.29 0.00
293 0.78 0.22 2 0 1.56 0.44 0.57 0.45 1.29 0.00
423 0.78 0.22 1 1 1.56 0.44 0.90 0.34 0.64 2.25
662 0.78 0.22 3 1 3.11 0.89 0.02 0.89 0.96 1.13
683 0.78 0.22 1 1 1.56 0.44 0.90 0.34 0.64 2.25
Republic 145 0.60 0.40 0 1 0.60 0.40 1.50 0.22 0.00 2.50
154 0.60 0.40 1 2 1.80 1.20 0.89 0.35 0.56 1.67
261 0.60 0.40 2 0 1.20 0.80 1.33 0.25 1.67 0.00
593 0.60 0.40 1 1 1.20 0.80 0.08 0.77 0.83 1.25
Total χ2 6.76 0.66
Pooled χ2 13 7 12.55 7.45 0.04 0.84 0.875 1.228
Heterogeneity χ2 6.71 0.57
52
Table 2.3. Continuous variables from habitat characteristics at kill sites of white-tailed
deer and mule deer investigated during 2002-2004 in northeastern Washington.
White-Tailed Deer Mule Deer
Feature n Mean (SD) n Mean (SD)
Mean Slope (degrees) 29 6.6 (4.9) 24 7.8 (4.6)
Mean Snow Depth (cm) 15 8.13 (11.3) 10 9 (14.7)
Mean Canopy Density (%) 30 80 (35.8) 25 77.3 (29.3)
Mean Shrub Density (%) 30 38.8 (37.2) 25 34.7 (31.3)
53
Table 2.4. Categorical variables from habitat characteristics at kill sites of white-tailed
deer and mule deer investigated during 2002-2004 in northeastern Washington.
White-Tailed Deer Mule Deer
Feature n n
Landform Class Forested Slope 21 15Open Slope 2 2
Stream Valley 5 4
Ridge Crest 2 2
Habitat Class Mixed Conifer 19 11Mixed Forest 8 11
Other 3 2
Aspect North & East 11 9South & West 19 12